Skip to main content
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Essays in Applied Microeconomics


This dissertation consists of three essays. All are in personnel economics, using data from the trucking industry. Training by firms is a central means by which workers accumulate human capital, yet firms may be reluctant to train if workers can quit and use their gained skills elsewhere. "Training contracts" that impose a penalty for premature quitting can help alleviate this inefficiency. The first essay from this dissertation studies training contracts in the U.S. trucking industry where they are widely used, focusing on data from one leading firm. Exploiting two plausibly exogenous contract changes that introduced penalties for quitting, I confirm that training contracts significantly reduce quitting. To analyze the optimal design of training contracts and their welfare consequences, I develop and estimate a structural learning model with heterogeneous beliefs that accounts for many key features of the data. The estimation combines weekly productivity data with weekly subjective productivity forecasts for each worker and reveals a pattern of persistent overconfidence whereby many workers believe they will achieve higher productivity than they actually attain. If workers are overconfident about their productivity at the firm relative to their outside option, they will be less likely to quit and more likely to sign training contracts. Counterfactual analysis shows that workers' estimated overconfidence increases firm profits by over $7,000 per truck, but reduces worker welfare by 1.5%. Banning training contracts decreases profits by $4,600 per truck and decreases retention by 25%, but increases worker welfare by 4%. Despite the positive effect of training contracts on profits, training may not be profitable unless some workers are overconfident.

A robust finding in experimental psychology and economics is that people tend to be overconfident about their ability. However, much less is known about whether overconfidence can be reduced or eliminated, particularly in field settings. The second essay of this dissertation provides new evidence using data from the workplace. A field experiment with a large trucking firm shows that workers tend to systematically overpredict their productivity and that their overconfidence is unaffected by whether workers receive financial incentives of different sizes for accurate guessing. Randomly informing workers about other workers' overconfidence reduces overconfidence in the short-run, but the effect fades within two weeks. Neither the incentives or information treatments have any effect on worker satisfaction or search behavior. Using long-term survey data from a second firm, I show that experience reduces overconfidence, but only quite slowly. Although workers at both firms exhibit aspects of Bayesian updating, overconfidence appears to be sticky and difficult to change.

The third essay analyzes worker referrals. Many firms use referrals in their recruitment and hiring procedures. Are these practices profitable, and if so, why? A model is developed where referrals may improve selection and reduce moral hazard. The model is tested

using extremely detailed personnel and survey data from a leading firm in the trucking industry. Referred workers are similar to non-referred workers across a large number of background characteristics and lab experimentally-measured dimensions of preferences. Referred workers are between 10-25% less likely to quit; the effects are strong across all groups of drivers, including new workers for whom the firm invests in expensive firm-sponsored general training. However, referred workers attain similar initial productivity and productivity growth as non-referred workers, and are no more likely to engage in various forms of moral hazard. The accumulation of friends after the starting work does not positively affect retention, productivity, or moral hazard. On net, the evidence is consistent with the idea that referrals benefit firms by selecting workers with a better fit for the job, as opposed to selecting workers with higher overall quality, by affecting worker behavior, or by changing job amenities.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View